Re: Looking for a real Fortran-66 compatible PC compiler (CP/M or DOS or at least also Windows, doesn't matter)

glen herrmannsfeldt <gah@ugcs.caltech.edu> wrote:


(big snip)


> [How could you forget FREQUENCY ? -John]


I do remember wondering not so many years ago why it wasn't
brought back again. People seem much more interested in
optimization that they were 50 years ago.


Getting out my copy (printed from a PDF somewhere on the web)
of the Fortran manual for the 704...


    "The FREQUENCY statement permits the programmer to give his
    estimate, for each branch-point of control, of hte frequencies
    with which the several branches will actually be executed
    in the object program."


That would be useful for static branch prediction, though maybe
dynamic prediction works so well that it isn't needed.


There is also a form estimating the number of cycles for a DO loop,
which would also presumably be useful for optimization.


-- glen
[There is a legend that FREQUENCY went away when someone realized that
a compiler had implemented it backwards and nobody noticed. One of
the reasons we use optimizers is that we trust them to write nasty
tricky code correctly that would be buggy and unmaintainable if we
tried to do it by hand. -John]

Return to the comp.compilers page.
Search the comp.compilers archives again.